Projects, not Proctors
How Rize-powered courses handle AI use.
The Problem
In the last few years, faculty have become accustomed to reading work that could have been written by anyone, knowing it was written by no one. In 2025, 88% of students used AI to complete assignments, and 94% of AI-written submissions went undetected.¹ ² Even failing, rushed essays were acts of contact between a mind and a page, and now that contact is being lost altogether.
You're asking me to go from point A to point B, why wouldn't I use a car to get there?
This was the protest of an NYU student when their professor challenged their use of AI.³ The remark is easy to dismiss as cynicism, but it reflects something possibly worse than cheating: a troubling calculation about whether the work in front of them is worth their own thinking. If AI can produce the output, then the output is not the hard part, and the assignment that demanded it has been reduced to an errand.
Online Proctored Exams as a Solution
For online courses, proctored exams present a natural solution. They promise higher, harder barriers to cheating and communicate to the student that the institution is taking dishonesty seriously. But those barriers come at the cost of student privacy and anxiety,⁴ and deepen the assumption that integrity is something institutions must impose rather than something students can own.
The evidence supporting whether they actually reduce cheating is unclear. A study of 1,760 students found that proctoring had no effect on the temptation to cheat but significantly increased test anxiety.⁴ And, much like AI detection, auto-proctoring suffers from an arms race: each new detection capability generates a new evasion technique, with no stable endpoint.⁵ The students who successfully cheat through proctoring tend to be the ones with greater resources and technical sophistication.⁵
Most importantly, and this applies to in-person proctored exams as well, they do not address the underlying question that is troubling students, and, more quietly, institutions: "why should I do this?". If students are to find long-term meaning in their learning, the answer cannot simply be: "because we are watching".
Rize's Approach: Answer the “Why?”
The right solution must raise the difficulty of dishonesty while answering that question. While there may be no uncheatable assessment, there is a set of conditions under which cheating becomes progressively less rational:⁶
- Perceived value of the task is high
- Difficulty of fabrication is high
- Pressure to perform is not too high
To influence all three, Rize-powered courses mirror real professional work, prioritizing projects as primary summative assessments, and quizzes as formative assessments. Rather than research essays, marketing students develop creative briefs for a product launch. Instead of a conceptual exam, Data Analytics students complete regular knowledge checks and then build recommendation models around real business problems of their choosing.
This is work professionals complete every day with full access to AI, but is still hard, and therefore human: knowing what to build, why, and being able to stand behind that reasoning when someone challenges it.
To see these real limitations of the tools themselves, Rize-powered courses often include “AI limitation” assignments where students are invited to use AI for a hard problem, and then assess the limitations. Students no longer ask “can’t an AI do this” when they have already answered that question themselves.
Each project is designed with the following:
- An original brief tied to the student's own career goals, so the work is not generic and the stake in it is personal
- Multiple weeks of iterative decisions and instructor feedback before the final deliverable, so students know their work will be seen, and critiqued and monitored at multiple stages
- A culminating video presentation where the student walks through their creative problem-solving and defends the decisions they made, shared in a public forum with their peers
- Supporting knowledge check quizzes that test knowledge-level recall, and help students with metacognition
Each element is designed to increase the perceived value of the work, create a clear connection between the purpose of the assessment and the student’s future, and make AI misuse as difficult and costly as possible.
For the student, the result is portfolio work they can take to an employer, share with friends and family, and be excited about. For the institution, it is the ability to assess deep learning outcomes that matter: judgment, strategic thinking, clear communication under pressure, and solving real-world problems.
These design elements do not make Rize projects uncheatable. They are designed so that cheating them costs the student something they would actually want to keep.
The Future
AI will keep improving, students will keep using it, and employers will expect them to.⁷
The question every institution now faces is what a credential actually certifies: what a student can recall under supervised conditions, or what they can build, solve, and defend. Employers will expect every graduate to drive the car from point A to point B. Those who succeed will be the ones who can go where the car cannot take them.
References
- Freeman, J. (2025). Student Generative AI Survey 2025. HEPI Policy Note 61. Higher Education Policy Institute. https://www.hepi.ac.uk/reports/student-generative-ai-survey-2025/
- Scarfe, P., Watcham, K., Clarke, A. & Roesch, E. (2024). "A real-world test of artificial intelligence infiltration of a university examinations system: A 'Turing Test' case study." PLOS ONE, 19(6). https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0305354
- Shirky, C. (2025). "Is AI Enhancing Education or Replacing It?" The Chronicle of Higher Education, April 29, 2025.
- Conijn, R., Kleingeld, A., Matzat, U. & Snijders, C. (2022). "The fear of big brother: The potential negative side-effects of proctored exams." Journal of Computer Assisted Learning, 38(6). https://onlinelibrary.wiley.com/doi/10.1111/jcal.12651
- Caines, A. (2021). "Rebuilding Student-Institution Trust after Online Proctoring." The Tambellini Group. https://www.thetambellinigroup.com/rebuilding-student-institution-trust-after-online-proctoring/
- Lang, J.M. (2013). Cheating Lessons: Learning from Academic Dishonesty. Harvard University Press.
- World Economic Forum (2025). Future of Jobs Report 2025. https://www.weforum.org/publications/the-future-of-jobs-report-2025/
Gregory's passion for his role at Rize is driven by the organization's commitment to making the positive aspects found in many colleges accessible to everyone. Beyond the professional sphere, Gregory finds enjoyment in a combination of endurance sports such as cycling, rowing, and running, as well as engaging in creative pursuits like painting, pottery, sewing.
Projects, not Proctors
How Rize-powered courses handle AI use.
The Problem
In the last few years, faculty have become accustomed to reading work that could have been written by anyone, knowing it was written by no one. In 2025, 88% of students used AI to complete assignments, and 94% of AI-written submissions went undetected.¹ ² Even failing, rushed essays were acts of contact between a mind and a page, and now that contact is being lost altogether.
You're asking me to go from point A to point B, why wouldn't I use a car to get there?
This was the protest of an NYU student when their professor challenged their use of AI.³ The remark is easy to dismiss as cynicism, but it reflects something possibly worse than cheating: a troubling calculation about whether the work in front of them is worth their own thinking. If AI can produce the output, then the output is not the hard part, and the assignment that demanded it has been reduced to an errand.
Online Proctored Exams as a Solution
For online courses, proctored exams present a natural solution. They promise higher, harder barriers to cheating and communicate to the student that the institution is taking dishonesty seriously. But those barriers come at the cost of student privacy and anxiety,⁴ and deepen the assumption that integrity is something institutions must impose rather than something students can own.
The evidence supporting whether they actually reduce cheating is unclear. A study of 1,760 students found that proctoring had no effect on the temptation to cheat but significantly increased test anxiety.⁴ And, much like AI detection, auto-proctoring suffers from an arms race: each new detection capability generates a new evasion technique, with no stable endpoint.⁵ The students who successfully cheat through proctoring tend to be the ones with greater resources and technical sophistication.⁵
Most importantly, and this applies to in-person proctored exams as well, they do not address the underlying question that is troubling students, and, more quietly, institutions: "why should I do this?". If students are to find long-term meaning in their learning, the answer cannot simply be: "because we are watching".
Rize's Approach: Answer the “Why?”
The right solution must raise the difficulty of dishonesty while answering that question. While there may be no uncheatable assessment, there is a set of conditions under which cheating becomes progressively less rational:⁶
- Perceived value of the task is high
- Difficulty of fabrication is high
- Pressure to perform is not too high
To influence all three, Rize-powered courses mirror real professional work, prioritizing projects as primary summative assessments, and quizzes as formative assessments. Rather than research essays, marketing students develop creative briefs for a product launch. Instead of a conceptual exam, Data Analytics students complete regular knowledge checks and then build recommendation models around real business problems of their choosing.
This is work professionals complete every day with full access to AI, but is still hard, and therefore human: knowing what to build, why, and being able to stand behind that reasoning when someone challenges it.
To see these real limitations of the tools themselves, Rize-powered courses often include “AI limitation” assignments where students are invited to use AI for a hard problem, and then assess the limitations. Students no longer ask “can’t an AI do this” when they have already answered that question themselves.
Each project is designed with the following:
- An original brief tied to the student's own career goals, so the work is not generic and the stake in it is personal
- Multiple weeks of iterative decisions and instructor feedback before the final deliverable, so students know their work will be seen, and critiqued and monitored at multiple stages
- A culminating video presentation where the student walks through their creative problem-solving and defends the decisions they made, shared in a public forum with their peers
- Supporting knowledge check quizzes that test knowledge-level recall, and help students with metacognition
Each element is designed to increase the perceived value of the work, create a clear connection between the purpose of the assessment and the student’s future, and make AI misuse as difficult and costly as possible.
For the student, the result is portfolio work they can take to an employer, share with friends and family, and be excited about. For the institution, it is the ability to assess deep learning outcomes that matter: judgment, strategic thinking, clear communication under pressure, and solving real-world problems.
These design elements do not make Rize projects uncheatable. They are designed so that cheating them costs the student something they would actually want to keep.
The Future
AI will keep improving, students will keep using it, and employers will expect them to.⁷
The question every institution now faces is what a credential actually certifies: what a student can recall under supervised conditions, or what they can build, solve, and defend. Employers will expect every graduate to drive the car from point A to point B. Those who succeed will be the ones who can go where the car cannot take them.
References
- Freeman, J. (2025). Student Generative AI Survey 2025. HEPI Policy Note 61. Higher Education Policy Institute. https://www.hepi.ac.uk/reports/student-generative-ai-survey-2025/
- Scarfe, P., Watcham, K., Clarke, A. & Roesch, E. (2024). "A real-world test of artificial intelligence infiltration of a university examinations system: A 'Turing Test' case study." PLOS ONE, 19(6). https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0305354
- Shirky, C. (2025). "Is AI Enhancing Education or Replacing It?" The Chronicle of Higher Education, April 29, 2025.
- Conijn, R., Kleingeld, A., Matzat, U. & Snijders, C. (2022). "The fear of big brother: The potential negative side-effects of proctored exams." Journal of Computer Assisted Learning, 38(6). https://onlinelibrary.wiley.com/doi/10.1111/jcal.12651
- Caines, A. (2021). "Rebuilding Student-Institution Trust after Online Proctoring." The Tambellini Group. https://www.thetambellinigroup.com/rebuilding-student-institution-trust-after-online-proctoring/
- Lang, J.M. (2013). Cheating Lessons: Learning from Academic Dishonesty. Harvard University Press.
- World Economic Forum (2025). Future of Jobs Report 2025. https://www.weforum.org/publications/the-future-of-jobs-report-2025/
Gregory's passion for his role at Rize is driven by the organization's commitment to making the positive aspects found in many colleges accessible to everyone. Beyond the professional sphere, Gregory finds enjoyment in a combination of endurance sports such as cycling, rowing, and running, as well as engaging in creative pursuits like painting, pottery, sewing.
